caches[create this folder yourself]：Store the processed data files that will be used for model training and testing, where in goalPlanning file stores the data generated by the goal planning module; responseGeneration saves the data generated by response generation module.
config：Store the pretrain model GPT2 and BERT, which are available for download from huggingface.
data：Training and test raw data that needs to be processed.
graph：The code for processing the data input from the heterogeneous graph and the code for updating the heterogeneous graph.
logs[create this folder yourself]：Store the trained model.
model：The training code and the model of goal planning.
outputs[create this folder yourself]：store the generated goal and response.
responseGeneration：The model and the data processing code of response generation module.
utils：Data processing code.
main.py：The entry file of goal planning.
GPU NVIDIA GeForce RTX 3090
Python 3.7
Pytorch 1.8.0
pip install -r requirments.txt
run main.py：
python main.py --mode train
You can change the parameters to train different datasets
python main.py --mode test
run run_train.py in the directory responseGeneration
cd responseGeneration
python run_train.py
run run_infer.py in the directory responseGeneration
cd responseGeneration
python run_infer.py
run eval_dialogue.py in the directory responseGeneration
cd responseGeneration
python eval_dialogue.py